arXiv:2610.07175v1 Announce Type: new Abstract: Visual latent reasoning compresses rendered derivations into compact intermediate states, reducing textual reasoning overhead. Existing approaches diff...
The paper introduces LLoCoT, a looped latent‑reasoning framework that replaces the traditional left‑to‑right generation of chain‑of‑thought tokens with iterative refinement of a compact latent workspace using a shared transformer. By jointly updating latent slots over a few iterations and sampling latent tokens in parallel, LLoCoT feeds these into an autoregressive decoder for answer generation. Experiments on HumanEval and MBPP show that LLoCoT matches or surpasses explicit‑CoT baselines while dramatically reducing reasoning latency and increasing throughput.
arXiv:2606. 16222v1 Announce Type: new Abstract: Large Language Models (LLMs) increasingly rely on intermediate reasoning, yet explicit Chain-of-Thought (CoT) suffers from a linguistic space bottleneck: each thought must be decoded into tokens, causing high inference overhead.
arXiv:2606. 06252v1 Announce Type: new Abstract: Recent work moves intermediate reasoning from natural-language traces into latent or cache-level representations to reduce token overhead and avoid a discrete communication bottleneck.
arXiv:2606. 06447v1 Announce Type: cross Abstract: Large language models often improve reasoning by generating explicit chain-of-thought (CoT), demonstrating the importance of intermediate computation.
The paper introduces Discrete Latent Reasoning (DLR), a method that transforms continuous latent states into explicit discrete tokens by rendering textual chains of thought into images, extracting visual features, and clustering them to build a discrete latent vocabulary. DLR enables standard autoregressive modeling over both natural language and latent tokens, supporting pretraining alignment, supervised fine-tuning, and reinforcement learning. Experiments on five reasoning benchmarks and two model series (Qwen3‑VL and LLaMA‑3) show that DLR outperforms prior latent reasoning baselines with up to 20× compression while preserving interpretable semantic structure.
arXiv:2607. 25915v1 Announce Type: new Abstract: Complex structured reasoning tasks often require additional computation, yet current language models obtain it mainly by increasing parameter scale or by serializing intermediate steps as chain-of-thought (CoT) tokens.
Large language models often improve reasoning by generating explicit chain-of-thought (CoT), demonstrating the importance of intermediate computation. However, textual CoT forces this computation through a discrete, serial, and communication-oriented token stream: each reasoning step must be verbalized before the model can proceed, even when the underlying update is semantic, uncertain, or only partially formed.
Lapras is a post‑training framework for Time Series Language Models that replaces explicit Chain‑of‑Thought reasoning with continuous latent thoughts in a joint time‑series‑language space. By self‑distilling from a teacher that generates textual rationales, the student model aligns its hidden states at the answer stage, transferring reasoning ability while producing only the final answer in text. Experiments on four TSLM backbones and five benchmarks show Lapras boosts average F1 by up to 10.79% over explicit CoT and reduces token generation by 23.9×, while still enabling readable reasoning traces through standard decoding.
arXiv:2605.06285v2 Announce Type: replace Abstract: Single-step retrieval-augmented generation (RAG) provides an efficient way to incorporate external information for simple question answering tasks...
The paper introduces Probabilistic Latent Reasoning Models (PLRM), a parallelizable variational framework that moves reasoning from token‑by‑token language generation into a continuous latent space. PLRM uses a JEPA‑style objective to infer latent distributions in parallel and a mixture‑based prior with a discrete strategy variable to preserve multiple distinct reasoning trajectories. Experiments show PLRM matches deterministic latent models on single‑trajectory tasks and outperforms stochastic baselines on multi‑trajectory benchmarks while enabling efficient training.
Flow matching enables language generation in few steps, but whether additional integration steps improve reasoning remains unclear. We prove that a flow parameterized by a two-layer Transformer can so...